2022 onwards sees no significant improvements in carbon dioxide pollution. Explore the map below to see for yourself:

Despite the prolific rise and alluring promise of large language-learning models in solving the climate crisis, their contributions to energy consumption and CO₂ emissions through computational costs and data storage have only worsened these issues. As exhibited by the data, the current trend of increasing pollution shows no signs of reversing—even with the advent of LLMs.

We, as data scientists, are uniquely positioned to address this challenge. It is our responsibility to either mitigate the impact of LLMs through sustainable training practices or to invent solutions that offset their carbon footprint. These tools must become assets in our climate strategy, not liabilities.

To learn more about what you can do as a data scientist to help the climate crisis, see Section 3 of Harvard Data Science Review's "Carbon Emissions in the Tailpipe of Generative AI"

U.S. Environmental Protection Agency Greenhouse Gas Reporting Program (GHGRP), 2010–2023.

This piece was inspired by James Temple's article "Sorry, AI Won't 'Fix' Climate Change" published by the MIT Technology Review.